EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[Refactor] [MoE] Rename moe-related classes & files (#3646)

Browse Source 
### What this PR does / why we need it?
1. Rename common_fused_moe.py to fused_moe.py.
2. Rename fused_moe_prepare_and_finalize.py / FusedMoEPrepareAndFinalize
to prepare_finalize.py / PrepareAndFinalize.
3. Rename vllm_ascend/ops/moe to vllm_ascend/ops/fused_moe.
4. Move vllm_ascend/ops/fused_moe.py to
vllm_ascend/ops/fused_moe/fused_moe.py
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
e2e & ut

- vLLM version: v0.11.0rc3
- vLLM main:
17c540a993

Signed-off-by: Pr0Wh1teGivee <calvin_zhu0210@outlook.com>
This commit is contained in:
weichen
2025-10-25 11:22:03 +08:00
committed by GitHub
parent 0637e8f021
commit 63c363d3de
25 changed files with 183 additions and 199 deletions
Download Patch File Download Diff File Expand all files Collapse all files

46
tests/ut/ops/test_moe_comm_method.py
View File

@@ -4,8 +4,9 @@ import torch
from vllm.model_executor.layers.fused_moe import FusedMoEConfig
from tests.ut.base import TestBase
from vllm_ascend.ops.moe.moe_comm_method import (AllGatherCommImpl,
AlltoAllCommImpl, MC2CommImpl)
from vllm_ascend.ops.fused_moe.moe_comm_method import (AllGatherCommImpl,
AlltoAllCommImpl,
MC2CommImpl)
class TestMoECommMethod(TestBase):
@@ -24,12 +25,14 @@ class TestMoECommMethod(TestBase):
self.moe_config.dp_group = MagicMock()
self.moe_config.num_global_redundant_experts = 0
@patch("vllm_ascend.ops.moe.moe_comm_method.get_current_vllm_config")
@patch("vllm_ascend.ops.moe.moe_comm_method.get_forward_context")
@patch("vllm_ascend.ops.fused_moe.moe_comm_method.get_current_vllm_config")
@patch("vllm_ascend.ops.fused_moe.moe_comm_method.get_forward_context")
@patch(
"vllm_ascend.ops.moe.moe_comm_method.FusedMoEPrepareAndFinalizeWithAllGather"
"vllm_ascend.ops.fused_moe.moe_comm_method.PrepareAndFinalizeWithAllGather"
)
@patch(
"vllm_ascend.ops.fused_moe.moe_comm_method.TokenDispatcherWithAllGather"
)
@patch("vllm_ascend.ops.moe.moe_comm_method.TokenDispatcherWithAllGather")
def test_all_gather_comm_impl(self, mock_token_dispatcher,
mock_prepare_finalize,
mock_get_forward_context,
@@ -72,12 +75,11 @@ class TestMoECommMethod(TestBase):
context_metadata=context_metadata)
mock_pf_instance.finalize.assert_called_once_with(h_out, True, None)
@patch("vllm_ascend.ops.moe.moe_comm_method.get_current_vllm_config")
@patch("vllm_ascend.ops.moe.moe_comm_method.get_forward_context")
@patch("vllm_ascend.ops.fused_moe.moe_comm_method.get_current_vllm_config")
@patch("vllm_ascend.ops.fused_moe.moe_comm_method.get_forward_context")
@patch(
"vllm_ascend.ops.moe.moe_comm_method.FusedMoEPrepareAndFinalizeWithMC2"
)
@patch("vllm_ascend.ops.moe.moe_comm_method.TokenDispatcherWithMC2")
"vllm_ascend.ops.fused_moe.moe_comm_method.PrepareAndFinalizeWithMC2")
@patch("vllm_ascend.ops.fused_moe.moe_comm_method.TokenDispatcherWithMC2")
def test_mc2_comm_impl(self, mock_token_dispatcher, mock_prepare_finalize,
mock_get_forward_context,
mock_get_current_vllm_config):
@@ -121,12 +123,14 @@ class TestMoECommMethod(TestBase):
context_metadata=context_metadata)
mock_pf_instance.finalize.assert_called_once_with(h_out, True, None)
@patch("vllm_ascend.ops.moe.moe_comm_method.get_current_vllm_config")
@patch("vllm_ascend.ops.moe.moe_comm_method.get_forward_context")
@patch("vllm_ascend.ops.fused_moe.moe_comm_method.get_current_vllm_config")
@patch("vllm_ascend.ops.fused_moe.moe_comm_method.get_forward_context")
@patch(
"vllm_ascend.ops.moe.moe_comm_method.FusedMoEPrepareAndFinalizeWithAll2All"
"vllm_ascend.ops.fused_moe.moe_comm_method.PrepareAndFinalizeWithAll2All"
)
@patch(
"vllm_ascend.ops.fused_moe.moe_comm_method.TokenDispatcherWithAll2AllV"
)
@patch("vllm_ascend.ops.moe.moe_comm_method.TokenDispatcherWithAll2AllV")
def test_alltoall_comm_impl(self, mock_token_dispatcher,
mock_prepare_finalize,
mock_get_forward_context,
@@ -163,13 +167,15 @@ class TestMoECommMethod(TestBase):
mock_pf_instance.prepare.assert_called_once_with(
hidden_states, router_logits, False, False, None)
@patch("vllm_ascend.ops.moe.moe_comm_method.get_current_vllm_config")
@patch("vllm_ascend.ops.moe.moe_comm_method.get_forward_context")
@patch("vllm_ascend.ops.fused_moe.moe_comm_method.get_current_vllm_config")
@patch("vllm_ascend.ops.fused_moe.moe_comm_method.get_forward_context")
@patch(
"vllm_ascend.ops.moe.moe_comm_method.FusedMoEPrepareAndFinalizeWithAllGather"
"vllm_ascend.ops.fused_moe.moe_comm_method.PrepareAndFinalizeWithAllGather"
)
@patch("vllm_ascend.ops.moe.moe_comm_method.TokenDispatcherWithAllGather")
@patch("vllm_ascend.ops.moe.moe_comm_method.unified_apply_mlp")
@patch(
"vllm_ascend.ops.fused_moe.moe_comm_method.TokenDispatcherWithAllGather"
)
@patch("vllm_ascend.ops.fused_moe.moe_comm_method.unified_apply_mlp")
def test_fused_experts_method(self, mock_unified_apply_mlp,
mock_token_dispatcher, mock_prepare_finalize,
mock_get_forward_context,